Submitter: C FP group / Fred Tydeman
Submission Date: 2018-08-01
Document: WG14 N2283
Reference Document: TS 18661-1
TS 18661-1 CFP CR 21
It is possible that a floating-point value, when converted to a one-digit character string, results in odd numbers no matter which way rounding is done. For roundTiesToEven, IEC 60559 specifies that the larger magnitude value be used.
Suggested Technical Corrigendum:
Add the following to TS 18661-1 clause 10.2 Conversions to character sequences:
At the end of F.5 Binary-decimal conversion, add:
NOTE IEC 60559 specifies that conversion to one-digit character strings using roundTiesToEven when both choices have an odd least significant digit, shall produce the value with the larger magnitude. For example, this can happen with 9.5e2 whose nearest neighbors are 9.e2 and 1.e3, both of which have a single odd digit in the significand part.